900 MHz PATH LOSS MEASUREMENTS AND PREDICTION TECHNIQUES FOR IN-BUILDING COMMUNICATION SYSTEM DESIGN. Scott Y. Seidel and Theodore S.

Transcription

1 ':\ 'J L v rc.-- qj {y\_~()d>qj 9 MHz PATH LOSS MEASUREMENTS AND PREDICTION TECHNIQUES FOR IN-BUILDING COMMUNICATION SYSTEM DESIGN Scott Y. Seidel and Theodore S. Rappaport Mobile and Portable Radio Research Group Bradley Department of Electrical Engineering Virginia Polytechnic Institute and State University Blacksburg, VA ABSTRACT This paper considers the effects of walls, floors, and building type on path loss in four buildings. Using a simple d" model, average path loss falls off to the n=3.14 power with a standard deviation of 16.3 db for the entire data set. However, when the effects of building type and number of floors between the transmitter and receiver are considered, standard deviations are reduced to 5 db about a mean d" power law. Average Floor Attenuation Factors (F AF) which describe the additional path loss caused by floors between the transmitter and receiver are found for as many as four floors in a typical office building. Preliminary evaluation of techniques to predict signal strength based on site-specific geometry are presented. I. INTRODUCTION An important step for the development and deployment of in-building high capacity wireless communications is understanding the propagation environment. Several researchers have measured radio waves in buildings and statistically modeled their results r1]-[8]. In this paper, we first present the statistical path loss results of 914 MHz measurements inside four buildings, and then attempt to explain the measured signals in more deterministic terms. First, path loss due to floor separation is found, and an improved path loss model having a simple form of d" is shown to substantially reduce the standard deviation about the mean. Then, we investigate propagation prediction in deterministic terms based on building blueprints, ray tracing, scattering, and diffraction theory. The buildings measured include a grocery store, a retail department store, and two multistory office buildings. Contour plots of locations of equal path loss for a fixed base transmitter are presented for two of the buildings. ll. MEASUREMENT PROCEDURE AND LOCATIONS Narrow band ( CW) signal strength measurements were made at 914 MHz with a system nearly identical to the one used in [3]. The stationary transmitter was placed at several locations within each of the buildings. For each transmitter location, the mobile receiver canvassed the building within measurement range of the transmitter. During each measurement, the mobile receiver moved at constant walking velocity along a straight path which varied in length between 2.4 and 6 meters, depending on surroundings. The open-plan shopping area of the grocery store, shown schematically in Figure 4 has dimensions of 46 meters by 67 meters and a ceiling height of 6 meters. This area consists of metal shelves that are 22 meters long and 2 meters high. The aisles next to the shelves are each 2.3 meters wide. The transmitter was located at a far side of the store while the receiver canvassed the entire shopping area. The single story open-plan retail store has inside dimensions of 61 meters by 52 meters and a ceiling height of 4.5 meters. The store is divided into several departments. Movable 1.8 meter high partitions (we call these soft partitions since they can be relocated) partially separate some of the departments, while other departments are divided by metal shelves ranging from meters high. Main aisles are 2 meters wide and secondary aisles are 1.2 meters wide. The three transmitter locations were near checkout lanes, in the lawn furniture department, and in the center of the store near small appliances. The receiver moved throughout the shopping area. The first office building has five floors and two wings where office areas are boxed in with soft partitions.. These partitions are cloth covered plastic dividers which divide each large open-plan office area into several smaller individual cubicles. Typical cubicles are 2.4 meters by 2.4 meters with soft partitions 1.5 meters tall. Aisles between the cubicles are about 1.2 meters wide. Figure 5 shows a schematic of the West wing of the 5th floor of office building 1. Office building 2 is a four story building with office areas that are very similar to those in office building 1. ill. DATA PROCESSING Figure 1 shows a typical measurement run where the receiver moved 12 meters along an aisle in the retail store. The abscissa represents the T-R separation, which only changes by 1 meters along the 12 meter measurement run since the receiver is not moving radially away from the transmitter. The median signal strength over a distance of 2.\ (6.56 m) was computed at 2..\ intervals for each measurement run and is represented as a point on a path loss scatter plot (see Figures 2-3). Thus each point represents the median path loss over a 6.56 meter section of a straight path. Median signal strengths were converted to absolute path loss by subtracting the median received signal strengths from the transmitted power of 29. dbm. This work has been supported by a National Science Foundation Graduate Research Fellowship and the MPRG Industrial Affiliates Program.

2 - 1 -s- 2 co ~- 3 ~- 4 ::) ~- 6 > a;_ 7 u Gl : T-R Separation <meters> Tx Ant : 1. 5 m Rx Ant : 1. 8 m Uel. :.413 m/s Figure 1 An example of CW signal fading as the receiver is moved in a retail department store. Fade depths which dip 3 db below the median can occur.. A model used in r1]-(6],[9] suggests that mean path loss increases exponentially with distance, that is Path Loss (d) oc (j;jn (1) where n is the path loss exponent which indicates how fast path loss increases with distance, d is a reference distance, and d is the T-R separation distance. When plotted on a log-log scale, this exponential relationship is a straight line. Absolute path loss in db is defined as the path loss in db from the transmitter to the reference distance d plus the additional path loss described by (1) in db. PL(d) db = PL(d ) + 1 x n x log 1~t) db (2) For these dat8'" a 1 meter reference distance was chosen and VJ'G <lb8'tu.ne P 1( d ) is d'l.lle to free space between the transmitter and tlie receiver located at 1 m. Measurements (i.e. (3]) show this is a valid assumption. Assuming antenna gains equal system cable losses, which is valid for our system, this leads to 31.7 db path loss at 914 MHz over a 1 m free space path. i.e., PL(l meter) db = 2 log 1 ( 4 11' (l meter)) db (3) We consider path loss to be log-normally distributed about the mean value power law described by (2) [1),(2]. Linear regression was used to compute values of n and (f, the standard deviation about the linear regression in db for the measured data. The effects of building type and the number of floors between the transmitter and receiver are included in the model to reduce the standard deviations and model the path loss more accurately as a function of t4e surroundings. IV. RESULTS A. Statistical.Mh Loss Mod.els Table 1 summarizes the mean path loss exponent values, and standard deviations about the means for different environments. Although the models presented here are for CW measurements, (2] showed that when individual path amplitudes are uncorrelated or phases of individual multipath com~onents are independent and identically distributed over l,211'), CW and wide band path loss measurements are equivafent when averaged over Table 1 n a(db) All Buildings : All Locations Same Floor Through 1 Floor Through 2 Floors Through 3 Floors Grocery store Retail store Office Building 1 : Entire Building Same Floor W. Wing 5th Floor Central Wing 5th W. Wing 4th Floor Office Building 2 : Entire Building Same Floor The mmtmum mean square error exponential path loss exponent and standard deviation for 914 MHz CW propagation for various combinations of the measured data. distances of a. few wavelengths. Thus, these models ma.y be used to describe average wide band path loss for these environments, as well. From Table 1, it can be seen that the path loss model for the entire data set is n=3.14 with a. large standard deviation of 16.3 db. This large value of (f is typical for data collec.ted from different building types. This model may be used for a. first-order prediction of mean signal strength when only T-R separation but no specific building information is known. For measurements 'It o " J1 M \'i'"l~~r~ ~Kt)~ ~)(.~~(.}il!j~.:t~li.l5tc:~ L~lu.u!\.::!~t-~~vc:! t:.:!~.:c~ tj!.!. t.!!..e ~~!!_I_~S!.!ll..u_.!. 1 u~2.7g and a~l2.9 dd ovm: tho xom.' buildiug9. The n value for all measurement runs in the grocery store is less than two. In the retail department store, mean path loss increases with distance slightly greater than free space and there is a relatively small spread about the mean value. The results for the grocery and retail stores, and thus probably the propagation measurements, strongly a~ree with those found in open-plan factory buildings (2J,(3),(5]. Scatter plots of path loss vs. T~R separation for the office building measurements are given in Figures 2 and 3. The dotted lines indicate the path loss models for n=1 through n=6 for a 1 meter reference distance. The dashed line indicates the best (minimum mean square error) mean path loss model for the data. presented in the graph. Different symbols are used to indicate data. from different environments, and overall n and (f are given on the left side of each graph. Multi-floor measurements were possible in the two office buildings, and nearly all measurements had multiple obstructions such as walls, windows, and soft partitions between the transmitter and receiver. From Figure 2, mean path loss increases with distance to the 3.54 power with a large standard deviation of 12.8 db. The simple d" path loss model in Flgure 2 does not use knowledge of office partitions or the number of floors between the transmitter and receiver. Transmissions between more floors lead to higher path loss. Notice the higher path loss when the receiver is in an

3 ~ 1 ---;; 9 (/).3 8..c 7 +-' /f Figure ,...,., 11 ~ 1 ---;; 9 (/).3 8..c o CW Path Loss- Office Building 1 n=6 Same Floor One Floor Two Floors Three Floors Four Floors Elewtor., "' f 91 4 MHz ~, ' : c ~ \illlj n 3.:54, o: ~. ~ Jl~ o OJ.. ~.....-a-. ~. ao ' :. '. 9- ~~~. ~ ' '=12.8d8 ': '' '.~ ~AJ~~- :. ~~: ~ ' '... '. '. n: ::... J n=5 n=4 n=3 n=2 n=1 1 1 Transmitter-Receiver Separation (m) Scatter plot of CW path loss as a function of distance in office building 1. The symbols represent the number of floors between the transmitter and receiver. a CW Path Loss - Office Building 2 n=6 Same Floor One Floor Two Floors Three Floors ' ''...,A' ~... Q-Da,... ffc;p~o O. f = 914 MHz I : 1 r!~. a 1).. ' ':.1/. 8J r::r. n "" 4.33 '. d a!p oo ~ ' = 1 J 3dB.. ' /~~. o~ 'S ' ' ' ' (' ' Q' (1:'. ', o o Jl.,. nrhr< :: ::...,.,. n=5 n=4 n=3 n=2 n=1 1 1 Transmitter-Receiver Separation (m) Figure 3 Scatter plot of CW path loss as a function of distance in office building 2. The symbols represent the number of floors between the transmitter and receiver. elevator than at locations with the same transmitterreceiver separation when the receiver is not in an elevator. The transmitter was in the basement and the receiver was in the elevator moving from the basement to the fifth floor. In office building 2, mean path loss increases with distance to the 4.33 power as shown in Figure 3. The number of floors between the transmitter and receiver can be seen to be an important parameter in the path loss model. All floors in the two office buildings were made of reinforced concrete, however, the attenuation through three floors in office building 1 is 24.4 db and the attenuation through three floors in office building 2 is 31.6 db. Office building 1 was built within the past ten years, and office building 2 is 2-3 years old. The standard deviations for individual buildings and for classification of measurements by number of floors in Table 1 are smaller than those for the entire data set. For example, the standard deviation for the entire data set. is 16.3 db. If we consider only office building 1, rr drops to 12.8 db. For same floor measurements only, rr=ll.2 db. Further classification of same floor transmitter and receiver locations into West wing 5th floor Central wing 5th floor, and West wing 4th floor reduce~ the standard deviations to 8.1 db, 4.3 db, and 4.4 db for each of the areas, respectively. The effect of the number of floors between transmitter and receiver may be modeled in at least two different ways. First, a different mean path loss exponent ( n) value which is a function of the number of floors may be used. These values are given in Table 1 for use in equation ( 4). Alternatively, a constant Floor Attenuation Factor (db) as a function of the number of floors may be added to the path loss modeled by a. same floor path loss exponent for the particular building (equation (5)). Values for the Floor Attenuation Factor in Table 2 are an average (in db) of the difference between the path loss observed at multi-floor locations and the path loss predicted by the simple d" model. n is the same floor exponent given in Table 1 for the particular building structure and d is the shortest distance, measured in 3 dimensions between the transmitter and receiver. PL( d) db=pl( d )+1. x n(multifloor) x log 1 (J) db ( 4) PL(d) db= PL(d )+1.xn(same floor) xlog 1 (t)+faf db (5) where dis in meters and PL(d ) =31.7 db. Table 2 Floor Attenuation Factor (db) Office Building 1 : Through 1 floor 12.9 Through 2 floors 18.7 Through 3 floors 24.4 Through 4 floors 27. Office Building 2 : Through 1 floor 16.2 Through 2 floors 27.5 Through 3 floors 31.6 Average Floor Attenuation Factor in db for one, two, three, and four floors between the transmitter and receiver in the two office buildings. B.s. Deterministic Path Loss Models A propagation prediction tool which could compute signal strength contours for a particular transm~tter location would be useful in in-building microcellular radio system design. The previous models include the effects of T-R separation, building type, and the number of floors between the transmitter and receiver, and are a first step to the ultimate goal of site-specific propagation prediction. Although standard deviations have been reduced, the standard deviations given in Table 1 for all single floor measurements in the two office buildings are still 11.2 db and 13.3 db. This indicates that transmitter and receiver locations relative to building features must be considered for a more accurate propagation model. This was done in [8}. Measurements at office building 1 were sometimes made with the transmitter and receiver in different wings

4 of the building. Table 1 indicates that when the transmitter and receiver are in different wings on the same floor in office building 1, the propagation is significantly different than if both the transmitter and receiver a.re in the same wing. The next section considers path lou models based on deterministic methods such as ray tracing, scattering, and diffraction theory which include the effects of specific building geometries in propagation. Blueprints and average path loss data. were imported to a computer-aided design program. The measured data. have been used to form contour plots of levels of equal path loss for a given transmitter location for the grocery store, and the fifth floor West wing of office building 1. Figure 4 is. the contour plot for the grocery store. The transmitter location is indicated by an 'X' on the left hand side of the drawing, and the long nearly rectangular objects represent typical 2.1 meter high grocery shelves. Curved solid lines indicate locations of equal path loss from the transmitter in 5 db steps. The amount of path loss is indicated in 1 db steps at the end of the lines. Consider the 6 db path loss line. Notice that at the center of the store, this contour is closer to the transmitter than at the edges. This confirms that shadowing by the many 2.1 meter high metal shelving units causes path loss to increase more rapidly than at locations where fewer shelving units a.re located between the transmitter and receiver. The contour plot for office building 1 is given in Figure 5. The transmitter is located in the upper right hand corner of the figure as indicated. Curved solid lines represent contours of equal path loss from the transmitter in 5 db steps. The thin lines on the drawing indicate 1.5 meter high cubicle dividers. In the center of the building are conference rooms with concrete block walls which span from the floor to the ceiling. These walls are indicated by thick lines on the drawing. In Figure 5, notice that when the thick conference room walls a.re between the receiver and the transmitter, the si~nal is attenuated much more rapidly than at other loca.t1ons. Since there is no direct line-of-sight path between the transmitter and receiver at these locations, energy must arrive at the receiver by other paths such as diffracted, scattered, or reflected paths. Notice that along the diagonal hallway along the edge of the building, the radio coverage is quite good, as it obeys better than free space propagation. This was also the case in [2),(3). Figure 6 shows the transmitter on the 5th floor in the West wing of office building 1. We attempt to model the path loss when a. receiver is in a lateral corridor such that there is no line-of-sight path. The transmitter and receiver are separated by 2.7 meters. We consider the four distinct ~ropagation paths shown in Figure 6: 1.) the direct path T-R), 2.) a single-hop path ( T-H-R), 3.) a scattered pat ( T S-R), and 4.) a. diffracted path ( T-D-R). For each case, we assume the entire received signal (therefore, path loss) is due to only one of the four propagation paths. First, consider the direct path ( T-R). If this path undergoes free space propagation, then the difference between measured and predicted path loss is 8. db (i.e., to explain the measurement, 8. db attenuation must be caused by the two concrete conference room walls between the transmitter and receiver). Next, assume that the only path is a single-hop reflection off the glass side wall ( T-H-R) with a reflection coefficient of -1 (perfect reflection). For the single-hop path to have equal angles of incidence and reflection at the reflecting boundary, this path must also pass through the concrete conference room walls near the receiver. Assuming the walls cause 8. db neters l Meters Figure 4 Contour plot of locations with equal path loss in 5 db steps for the grocery store.

5 Meters s~~eters ~ Figure 5 Contour plot of locations with equal path loss in 5 db steps for the 5th floor of the West wing of office building 1. Figure 6 Location in office building 1 where we attempt to model the path loss at R by the direct path ( T-R), a single-hop path ( T-H-R), a scattered path ( T-S-R), and a. diffracted path ( T-D-R). loss, the perfectly reflected path predicts a. path loss within 1 db of the measured path loss. If we allow the possibility of unequal angles of incidence and reflection a.t locations along the side wall, we ca.n examine a. scattered path ( T-S-R) where the sca.tterer has some radar cross-section and the hi-static radar equation [14) can be used to find path loss. For the measured and predicted path loss to be equal, ~he ra.da.r cross-section of the wall must be + 17 dbm. The scattered path has a. 4o jncidence angle, and a. 32 reflection angle. Hence, the scattered path alone is not adequate to predict the path loss at the receiver. The fourth case to consider is diffraction between the transmitter a.nd receiver along the path ( T-D-R). The corner of the conference room is modeled as a knife-edge at D. Simple knife-edge diffraction theory (11 predicts a total path loss of 74.2 db, which is 8.2 db weaker than measured. Thus, the diffracted path alone cannot accurately predict the measured path loss. For the specific case considered, it is apparent that no single ra.y modeling technique is adequate to predict the path loss where there is no line-of-sight path between the transmitter and receiver. However, when a.ll propagation mechanisms are considered in conjunction, it may be possible to predict path loss in these obstructed regions. For example, when a. two ray model which includes the relative phase of each path is used to combine the direct and reflected paths ( T-R) and ( T-H-R), the predicted and measured path loss agree to within 1 db if we assume the concrete walls attenuate each path by 8. db. The critical elements of each propagation mechanism such as attenuation through walls, reflection coefficients, and hi-static radar cross-sections, must. be modeled individually from models or measurements where the paths can be separated through the use of a. wide band probe [2),[5)-f7J. With amplitude and time delay information due to multiple paths, it is likely that the path loss can be accurately modeled. Figure 7 shows measured and predicted signal strength as the receiver moves along the diagonal hallway along the edge of office building 1. The abscissa indicates the T-R separation. The smooth curve is the predicted signal stren~th for the simple plane-earth two-ray model used in (1),lll] which assumes a direct ray and a reflected

6 - 1 "- e 2 ~- 3 ~- 4 :J a:- 6 > a;_ 7 u Ql o:_ ol7.o T-R Separation (meters) Tx Ant : 1. 5 m Rx Ant : 1. 8 m Uel. :.965 m/s Figure 7 Measured CW signal strength along a. hallway in the 5th floor West wing. The receiver has linem ofmsight to the transmitter. Also shown is the signal strength predicted by a simple two-ray plane earth propagation model for the given geometry. ray due to a floor (ground) bounce with reflection coefficient of -1. The two ray model accurately predict the dips at 9 m and 16 m T-R separations as seen in Figure 7. Additional rays due to wall, ceiling, and other reflections can likely be added to more accurately model the signal strength and produce additional fades. The agreement between the measured signal and the simple two-ray model in Figure 7 is encouraging. Indeed, two and four ray models have been shown to be good models for outdoor microcellular environments (12),(13). With the incorporation of the many additional rays that are present!!! ~.n!!ld'dn!'!!!~!lti~~th e!!"!i~~!!!!!ent f!:\1. it ig p~~~ib!e th~~ [t wholly detgl\'l':k~hiis~ic ~'c,yab'.'d.dng solf wvj.'c p~li.'ogr~nn th11t predicts average signal strength throughout a building can become a reality in the near future. V. CONCLUSIONS Narrow band path loss models at 914 MHz for four different buildings have been presented. The models are based on a simple dn exponential path loss vs. distance relationship. The models have been shown to be more accurate. when different b11ildings and dissimilar areas within the same building are considered separately. A floor dependent path loss exponent (Table 1) may be used to model the effects of the number of floors between the transmitter and receiver. Alternatively, the path loss exponent for co-floor propagation along with a Floor Attenuation Factor (Table 2) to account for the additional path loss due to floors, may be used to predict signal strength. Formulas for these two path loss models are given in the paper. Contour plots of regions of equal path loss have been presented for two buildings. These plots show where shadowed regions exist for a given transmitter location. In office building 1, the two-ray plane-earth model indicates that ray tracing may be used to predict signal strength in some LOS locations. To predict. signal strengths in all locations, ray tracing, scattering, diffraction, and shadowing should all be considered as potential propagation mechanisms. ACKNOWLEDGEMENTS The authors would like to thank Mike Keitz and Ken Blackard for their help in data collection. REFERENCES (1] D.C. Cox, R.R. Murray, and A.W. Norris, "8 MHz Attenuation Measured In and Around Suburban Houses," AT&T BLTJ, Vol. 63, July/August, 1984, pp [2] T.S. Rappaport, "Characterization of UHF Multipath Radio Channels in Factory Buildings," IEEE Transactions on Antennas and Propagation, Vol. 37, No. 8, August 1989, pp (3] T.S. Rappaport, and C.D. McGillem, "UHF Fading in Factories," IEEE Journal on Selected Areas in Communications, Vol. 7, No. 1, January 1989, pp (4] F.C. Owen, and C.D. Pudney, "Radio Propagation for Digital Cordless Telephone at 17 MHz and 9 MHz," Electronics Letters, Vol. 25, No. 1, January 5, 1989, pp. 52M53. [5] D.A. Hawbaker, and T.S. Rappaport, "Indoor Wideband Radiowave Propagation Measurements at 1.3 GHz and 4. GHz," Electronics Letters, Vol. 26, No. 21, October 11, 199, pp (6] A.A.M. Saleh, and R.A. Valenzuela, "A Statistical Model for Indoor Multipath Propagation," IEEE Journal on Selected Areas in Communications, Vol. SAC-5, No. 2, February, 1987, pp (7] R. Ganesh, and K. P ahla.van, "On the Modeling of Fading Multipath Indoor Radio Channels," IEEE Global Communications Conference, Dallas, TX, November, 1989, pp (8] J-F. Lafortune, and M. Lecours, "Measurement and 1\A'~..-1~1!~,. ~t n.. ~nn=n~:~~ T ~nn~n :~ ~ ' :1,1:~~ ~~ onn... vl!.\-1'~~~1\.~.ti.~j'.. '\II. A A'-Jifo.I~J~.I-'..'-~JliA\..JA!'...to-e~V~Ju~.; AAA t..l!.o ~L~AA\_-!\.l\..1!,-A.~ -~ --' L~\f\1 MHz, 11 IEEE 'l'?:a,'i'sactions on Vehicula't 'l'echnology, Vol. 39, No. 2, May 199, pp (9] T.S. Rappaport, S.Y. Seidel, and K. Takamizawa, "Statistical Channel Impulse Response Models for Factory and Open. Plan Building Radio Communication System Design," IEEE Transactions on Communications, to be Published April, 1991'. (1] J.D. Parsons, and J.G. Gardiner, Mobile Communication Systems, Blackie, London, 1989, pp [11] T.S. Rappaport, and L.B. Milstein, "Effects of Path Loss and Fringe User Distribution on CDMA Cellular Frequency Reuse Efficiency," IEEE Global Communications Conference, San Diego, CA, December 199, pp (12] A.J. Rustako, et. al., "Radio Propagation Measurements At Microwave Frequencies for Microcellular Mobile and Personal Communications," 1989 IEEE International Communications Conference, Boston, pp (13] E. Green, "Radio Link Design for Microcellular Systems," British Telecom Technology Journal, Special Issue on Mobile Communications, Vol. 8, No. 1, January 199, pp (14] M.I. Skolnik, Introduction to Radar Systems, McGraw-Hill, New York, 198, p. 557.